Author: Maciej Fijalkowski <[email protected]>
Branch: extradoc
Changeset: r3785:b7f86c9c1884
Date: 2011-06-24 10:05 +0200
http://bitbucket.org/pypy/extradoc/changeset/b7f86c9c1884/

Log:    merge

diff --git a/sprintinfo/genova-pegli-2011/directions.txt 
b/sprintinfo/genova-pegli-2011/directions.txt
new file mode 100644
--- /dev/null
+++ b/sprintinfo/genova-pegli-2011/directions.txt
@@ -0,0 +1,38 @@
+How to go to Genova Pegli
+=========================
+
+By train
+--------
+
+- http://www.trenitalia.com
+
+- Take a long distance train to Genova Piazza Principe or Genova Brignole
+  (both works; in case of doubt, pick Genova Principe as it's slightly closer
+  to Pegli)
+
+- From there, take a regional train to Genova Pegli: take one whose final
+  destination is Genova Voltri, Savona or Ventimiglia.  Beware that not all of
+  those actually stops in Pegli, so make sure that yours does :-) (in case of
+  doubt, you can ask a random person on the platform, they'll know it for
+  sure)
+
+- You can search for the timetable at the trenitalia.com website
+
+- This is the map from the Genova Pegli station to the Hotel: 
http://maps.google.it/maps?saddr=Genova+Pegli&daddr=Lungomare+di+Pegli,+22,+16155+Genova+(Albergo+Puppo)&hl=it&sll=44.42542,8.81594&sspn=0.001927,0.003793&geocode=FVrkpQId9oeGACllN1h7SD_TEjEhQe02_AQZnQ%3BFYDdpQIdaYGGACHNe85zd7hOuykraHuSRz_TEjHnjlgjZyCfOA&mra=ltm&dirflg=w&z=18
+
+
+By plane
+--------
+
+- http://www.airport.genova.it/v2/
+
+- From the airport, take the "Volabus" until the stop "Via Cornigliano /
+  Stazione FS":
+  
http://www.airport.genova.it/v2/index.php?option=com_content&view=article&id=67&Itemid=136&lang=en
+
+- From the Genova Cornigliano train station, take a regional train to Genova
+  Pegli whose final destination is Genova Voltri, Savona or Ventimiglia.  You
+  can use the same ticket as for the Volabus
+
+- Look at the map above for the hotel
+
diff --git a/talk/iwtc11/paper.tex b/talk/iwtc11/paper.tex
--- a/talk/iwtc11/paper.tex
+++ b/talk/iwtc11/paper.tex
@@ -808,8 +808,11 @@
 jump($L_1$, $p_{0}$, $i_8$)
 \end{lstlisting}
 
-XXX explain that this is effectively type-specializing a loop
-
+If all the optimizations presented above are applied, the resulting
+optimized peeled loop will consist of a single integer addition
+only. That is it will become type-specialized to the types of the
+variables \lstinline{step} and \lstinline{y}, and the overhead of
+using boxed values is removed.
 
 \section{Benchmarks}
 
@@ -825,7 +828,6 @@
 chose to present benchmarks of small numeric kernels where loop peeling can 
show
 its use.
 
-XXX we either need to explain that we use C++ or consistently use C
 \begin{figure}
 \begin{center}
 {\smaller
@@ -838,7 +840,7 @@
 \hline
 conv3(1e6) & 77.15 & 9.58 & 1.69  & 0.77 &  0.74 \\
 \hline
-conv3x3(1000) & 23.72 & 12.77 & 0.07  & 0.05 &  0.25 \\
+conv3x3(1000) & 236.96 & 128.88 & 0.70  & 0.41 &  0.25 \\
 \hline
 conv3x3(3) & 23.85 & 12.77 & 0.10  & 0.07  &  0.27 \\
 \hline
@@ -848,7 +850,7 @@
 \hline
 dilate3x3(1000) & 23.29 & 12.99 & 0.41  & 0.39 &  0.26 \\
 \hline
-sobel(1000) & - & - & - & - & 0.20 \\
+sobel(1000) & 181.49 & 95.05 & 0.71 & 0.42 & 0.20 \\
 \hline
 sqrt(Fix16) & 744.35 & 421.65 & 3.93  & 2.14  & 0.96 \\
 \hline
@@ -863,7 +865,11 @@
 }
 \end{center}
 \label{fig:benchmarks}
-\caption{Benchmark Results in Seconds}
+\caption{Benchmark Results in Seconds. Arrays of length $10^5$ and
+  $10^6$ and matrixes of size $1000\times 1000$ and $1000000 \times
+  3$ are used. The one used in each benchmark is indicated in
+  the leftmost column. For the matrixes, only the number of rows are
+  specified.} 
 \end{figure}
 
 \subsection{Python}
@@ -897,10 +903,12 @@
 \end{itemize}
 
 The sobel and conv3x3 benchmarks are implemented
-on top of a custom two-dimensional array class,  Array2D.
+on top of a custom two-dimensional array class.
 It is
 a simple straight forward implementation providing 2 dimensionall
-indexing with out of bounds checks. 
+indexing with out of bounds checks. For the C implementations it is
+implemented as a C++ class. The other benchmarks are implemented in
+plain C. 
 
 Benchmarks were run on Intel i7 M620 @2.67GHz with 4M cache and 8G of RAM in
 32bit mode.
_______________________________________________
pypy-commit mailing list
[email protected]
http://mail.python.org/mailman/listinfo/pypy-commit

Reply via email to